Python pandas groupby 按类别占总数的百分比

Question

I have following table:

+-----+----------+---+
| Grp | Category | X |
+-----+----------+---+
|   1 | A        | 1 |
|   1 | B        | 3 |
|   1 | B        | 2 |
|   1 | C        | 2 |
|   2 | A        | 2 |
|   2 | A        | 4 |
|   2 | B        | 4 |
|   3 | A        | 3 |
|   3 | C        | 7 |
+-----+----------+---+

And trying to get following:

+-----+----------+---------+
| Grp | Category | X_ratio |
+-----+----------+---------+
|   1 | A        | 1/8     |
|   1 | B        | 5/8     |
|   1 | C        | 2/8     |
|   2 | A        | 6/10    |
|   2 | B        | 4/10    |
|   3 | A        | 3/10    |
|   3 | C        | 7/10    |
+-----+----------+---------+

And a little bit stuck. Coud anyone suggest efficient solution?

My current code - its working, but seems not efficient:

grp_Cat = df.groupby(['Grp', 'Category ']).agg({'X': 'sum'})
grp_total = df.groupby(['Grp']).agg({'X': 'sum'})
grp_Cat.div(grp_total, level='Grp') * 100

Answer 1

Because performance is important first aggregate sum to MultiIndex Series and then divide by Series.div summed values per first Grp level:

s = df.groupby(['Grp','Category'])['X'].sum()
df = s.div(s.sum(level=0), level=0).reset_index(name='X_ratio')
print (df)
   Grp Category  X_ratio
0    1        A    0.125
1    1        B    0.625
2    1        C    0.250
3    2        A    0.600
4    2        B    0.400
5    3        A    0.300
6    3        C    0.700

Slowier alternative:

df = (df.groupby(['Grp','Category'])['X'].sum()
        .groupby(level=0)
        .apply(lambda x: x / x.sum())
        .reset_index(name='X_ratio'))
print (df)
   Grp Category  X_ratio
0    1        A    0.125
1    1        B    0.625
2    1        C    0.250
3    2        A    0.600
4    2        B    0.400
5    3        A    0.300
6    3        C    0.700